Migration of Credit Risk Models from SAS to Python

01The Client

Large financial institution with a dedicated Credit Risk Analysis department, with an analytical modeling environment consolidated on proprietary SAS technology.

02The Challenge

• Dependency on proprietary SAS technology with high licensing costs
• Difficulty in scaling and maintaining models in the legacy environment
• Limitation in adopting modern Machine Learning libraries
• Need to modernize the modeling environment to increase agility and reduce operational costs

03Implemented Solution

• Migration of credit risk analytical models from SAS to Python with Spark
• Rewriting and statistical comparative validation of models to ensure equivalence of results
• Implementation of distributed processing pipelines with PySpark
• Adoption of modern ML libraries (scikit-learn, XGBoost) for predictive credit modeling

04Strategic Differentials

• Elimination of proprietary software dependency with migration to open source stack
• Rigorous validation of original business logic during transition, without loss of model reliability
• Scalable architecture with Spark to support growing volumes of credit data
• Internal team training in modern Data Science and ML technologies

05Results Achieved

• Reduction in operational costs by eliminating licenses
• Greater agility in developing, updating and deploying risk models
• Access to open source ML ecosystem for continuous model evolution
• Scalability to process growing data volumes without additional licensing costs